Within the IoT world there has been an increasing buzz around the concept of ‘Physical AI’, the idea that IoT deployments will act as the eyes and ears of Artificial Intelligence as well as extending AI-based decision-making and actuating into the physical world. In June 2026, Transforma Insights published a report ‘AI and IoT: what are the implications of the convergence of the technology domains?’ looking at how the increasing adoption of AI within IoT will drive adoption and require new approaches to all aspects of how IoT is managed. In this article we explore a number of the key trends that will be stimulated by these changes.
Extending AI into the real world will drive IoT adoption
Transforma Insights takes the view that this is potentially one of the most valuable applications of AI, i.e. the potential efficiency savings and other benefits are most pronounced at the intersection of the physical and digital worlds, which is really where IoT sits. For instance, some of the most demonstrable impacts of AI would come with use cases like autonomous driving, or on a more mundane level the efficiency (in terms of cost, energy, fuel, wellbeing and more) savings from greater efficiency in operations like defect detection, workflow optimisation, fleet route planning, PPE detection and many more. AI-enabled IoT has the potential to create more tangible value than many purely digital AI applications, as well as reducing the cost of deployment.
The implicit expectation is for lots more demand for IoT deployments, either because AI facilitates more efficient data processing – and thus justifies deployments that might otherwise not be viable – or creates the demand for ever more data to be processed.
Key use cases
The potential use cases are extensive, across almost all types of IoT deployments. Some examples are presented below:
Predictive Maintenance in Manufacturing
Industrial machines generate continuous streams of sensor data (temperature, vibration, pressure, acoustics). AI can analyse millions of readings in real time to predict equipment failures before they occur. Without AI, organisations often collect the data but struggle to extract actionable insights. The ability to reduce downtime and maintenance costs creates a strong business case for more sensors.
Smart Grid Energy Management
Electric utilities receive data from millions of smart meters, transformers, and grid sensors. AI can rapidly detect demand patterns, forecast consumption and identify outages. This makes large-scale IoT deployments economically viable because utilities can act on the data instead of merely storing it.
Connected Vehicle Fleets
Trucks, delivery vans and commercial vehicles generate GPS, engine, fuel and driver-behaviour data continuously. AI can optimise routes, predict maintenance needs, and improve fuel efficiency across entire fleets. The value generated from these insights encourages broader adoption of vehicle telematics and IoT devices.
Smart Buildings
Modern buildings may contain thousands of sensors monitoring occupancy, lighting, HVAC systems, and security. AI can process these data streams simultaneously to reduce energy consumption and improve occupant comfort. Organisations are more likely to invest in building IoT infrastructure when AI converts sensor data into measurable savings.
Precision Agriculture
Farms deploy soil moisture sensors, weather stations, drones and irrigation monitors. AI can combine and analyse these diverse datasets to optimise irrigation, fertiliser use and crop management. Faster and more accurate decision-making increases the return on investment for agricultural IoT deployments.
Healthcare Remote Monitoring
Wearables and medical devices generate continuous patient data such as heart rate, blood oxygen levels and activity metrics. AI can rapidly identify anomalies that may indicate emerging health issues. Healthcare providers gain more value from connected devices because they can focus on high-risk patients rather than manually reviewing all data.
Smart Cities and Traffic Management
Traffic cameras, road sensors, parking systems, and public transportation networks generate enormous amounts of data. AI can analyse conditions in real time to optimise traffic signals, reduce congestion and improve public services. Demonstrable operational improvements justify expanding city-wide IoT sensor networks.
Retail and Supply Chain Visibility
RFID tags, warehouse sensors, inventory trackers and logistics devices continuously report status and location data. AI can detect bottlenecks, forecast inventory needs, and identify supply chain disruptions much faster than traditional analytics. As businesses see tangible operational benefits, they are more willing to deploy additional IoT devices throughout the supply chain.
But this is not a slam dunk
In the report, Transforma Insights’ analysts explore some of the major implications of the trend, for instance in terms of the increasing need for edge processing, the management of models across edge and cloud locations, the requirement for new software platforms, and the changing dynamics for connectivity.
The conclusion is that there are numerous substantial challenges associated with integrating AI into IoT. In the report we present them as eight ‘speed bumps’:
- Device heterogeneity and fragmentation: A diverse estate of devices with varying compute, memory, connectivity and operating environments complicates deployment, optimisation, monitoring, and support, increasing operational complexity and reducing standardisation opportunities.
- Resource constraints and fault tolerance: Many edge devices operate with limited processing power, memory, storage and energy availability. These constraints require careful optimisation while maintaining resilience during intermittent connectivity and infrastructure disruptions.
- Balancing real-time performance and accuracy: Achieving low-latency inference while maintaining acceptable model accuracy can be challenging, particularly when network conditions, hardware capabilities, and operational environments vary significantly across deployments.
- Model lifecycle management at scale: Managing model deployment, testing, validation, updates, version control, monitoring and rollback across large numbers of distributed devices introduces significant operational and governance challenges.
- Expanded security attack surface: Embedding AI into IoT systems creates new attack vectors, including model theft, adversarial manipulation, data poisoning, inference tampering and compromise of deployment and update mechanisms.
- Compliance, governance and privacy risks: Organisations must ensure AI-enabled IoT deployments comply with evolving regulatory requirements while protecting sensitive data and preventing unintended exposure of personal or operational information.
- Business process adaptation: Realising value from AI-enabled IoT often requires redesigning workflows, decision-making processes, operating models, and performance metrics to take advantage of new automation capabilities. These challenges were implicit in IoT as a stand-alone, and are at least equally pronounced with converged AI/IoT.
- Increased multidisciplinary complexity: AI adds new technical disciplines, tools, governance requirements, and skills to an already complex IoT domain spanning hardware, software, networking, cybersecurity, operations and data management. One of the main inhibiting factors on IoT is complexity inertia (i.e. resolving all the hardware, software and connectivity challenges), a trend which is even more relevant when AI is added to the mix.
Success in combining AI with IoT is dependent on overcoming these challenges.
Article by Matt Hatton, a founding partner at Transforma Insight
Comment on this article via X: @IoTNow_
